Re: [PATCH 2/3] rpmsg: Only invoke announce_create for rpdev with endpoints

[Bjorn Andersson]

On Tue 03 Apr 02:12 PDT 2018, Loic PALLARDY wrote:
> > -----Original Message-----
> > From: linux-remoteproc-owner@xxxxxxxxxxxxxxx [mailto:linux-remoteproc-
> > owner@xxxxxxxxxxxxxxx] On Behalf Of Bjorn Andersson
[..]
> > diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c
> > index dffa3aab7178..e85d2691d2cf 100644
> > --- a/drivers/rpmsg/rpmsg_core.c
> > +++ b/drivers/rpmsg/rpmsg_core.c
> > @@ -442,7 +442,7 @@ static int rpmsg_dev_probe(struct device *dev)
> >  		goto out;
> >  	}
> > 
> > -	if (rpdev->ops->announce_create)
> > +	if (ept && rpdev->ops->announce_create)
> 
> This check is already part of virtio_rpmsg.c (see line 341)
> 	/* need to tell remote processor's name service about this channel ? */
> 	if (rpdev->announce && rpdev->ept &&
> 	    virtio_has_feature(vrp->vdev, VIRTIO_RPMSG_F_NS)) {

The introduction of rpdev->ept in this check was done by Henri, as he
was implementing rpmsg_char support for virtio_rpmsg:

b2599ebffb2d ("rpmsg: virtio_rpmsg_bus: fix announce for devices without endpoint")

It's there because the rpmsg_device will not have a primary endpoint and
as such there's no communication channel to announce. We could add this
check in each implementation of announce, but I think it's a common
issue.

> 
> should it be part of qcom_smd driver too? (but each implementation will duplicate checks)
> Or may have a generic check in the core including rpdev->announce as well (and doing virtio_rpmsg.c clean-up)
> 

I think we want to remove the ept check in virtio_rpmsg, to reduce that
part of the duplication at least.

> Change will become:
> 	if (rpdev->announce && ept && rpdev->ops->announce_create)
> 

That might make sense, let's see what Wendy comes up with related to
rpmsg_char and supporting Linux-side services, as I suspect that this
handling might be affected.

Regards,
Bjorn